Short, well-written examples grouped by job rather than by novelty. The useful thing about an official library is that it reflects what the model was actually tuned for.
Prompt libraries worth copying from
Published prompt collections, and the tools that help you find out whether a prompt actually works. The second part matters more than the first.
Runnable notebooks rather than a prompt list, which makes it the better reference when the question is how something should be structured rather than what to type.
The closest thing to a textbook on the subject, with the citations to back it up. Read the techniques section and skip the model-specific parts.
Learn Prompting
Learn PromptingStructured like a course, from the basics to security. More patient than most of the field and better for someone starting from nothing.
Awesome ChatGPT Prompts
CommunityThe repository that made the role-prompt format popular. Enormous, uneven, and still the fastest place to find a starting point for a familiar task.
LangChain Hub
LangChainPrompts as versioned artefacts you pull into code, which is the right shape once a prompt is part of a product rather than a conversation.
PromptLayer
PromptLayerRecords what was sent and what came back, so a regression has a diff. The unglamorous tool most teams need before they need anything cleverer.
promptfoo
Open sourceTurns prompts into test cases you can run in CI, which is the only way a prompt library stays honest once more than one person edits it.
Prompt flow
MicrosoftA visual builder for prompt pipelines inside Azure, with evaluation wired in. Worth it if you are already there, heavy if you are not.
Humanloop
HumanloopEvaluation plus human review in one place, aimed at teams where a domain expert has to sign off on the output rather than an engineer.
FlowGPT
FlowGPTA consumer marketplace of shared prompts and characters. Useful for browsing what people actually use; the quality ceiling is set by what is popular rather than what works.
Agenta
Open sourceSelf-hostable prompt management with evaluation, which matters when the prompts touch anything you cannot paste into a vendor dashboard.
Latitude
LatitudeBuilt around catching the cases where the model goes off the rails rather than the average case, which is where the time actually goes.
PromptBase
PromptBaseA marketplace where prompts are sold, mostly for image generation. Worth knowing it exists; the free libraries cover most of what is on sale.
PromptHero
PromptHeroSearch over prompts paired with the images they produced, which is the only review mechanism image prompting has ever had.
Gemini prompting guide
GoogleGoogle own guidance, and the most explicit of the big three about which techniques their model responds to. Read it alongside the model, not instead of it.
Snack Prompt
Snack PromptA community library with an upvote mechanic, which at least gives you a weak signal about which prompts survived contact with real use.
Short and hands-on, structured around the failure modes rather than the features. The fastest of the official courses to work through.
PromptPerfect
Jina AIRewrites a prompt for you. Occasionally useful as a second opinion, and a bad habit as a first step: you learn less each time it works.
OpenAI prompting guide
OpenAIThe official reference, and the one to check first when a technique you read about elsewhere turns out not to apply to the model you are using.
Sources and method
The ranking is ours. Anything on this page that can be checked is checked against a named source, and the date we last opened that source is printed with it. If a figure and our verdict disagree, trust the figure and follow the link.
A model is listed only once its provider publishes pricing and documents an API. Every figure carries a source and the date we checked it; the ranking is ours.